While transcription factors play a critical role in orchestrating the gene expression profiles of all organisms, other, "epigenetic" levels of information that encode the diversification program of an otherwise uniform. genetic content exist. Methylation of DNA is thought to be one such critical determinant of the diversification program (Razin et al., 1980, Science 210:604-610).
DNA methylation is a postreplicative covalent modification of DNA that is catalyzed by the DNA methyltransferase enzyme (MeTase) (Koomar et al., 1994, Nucl. Acids Res. 22:1-10; and Bestor et al., 1988, J. Mol. Biol. 203:971-983). In vertebrates, the cytosine moiety at a fraction of the CpG sequences is methylated (60-80%) in a nonrandom manner generating a pattern of methylation that is gene and tissue specific (Yisraeli and M. Szyf, 1985, In DNA methylation: Biochemistry and Biological significance, pp. 353-378, Razin et al., (Ed), Springer-Verlag, N.Y.). It is generally believed that methylation in regulatory regions of a gene is correlated with a repressed state of the gene (Yisraeli and Szyf, 1985, In DNA methylation: Biochemistry and Biological significance, pp. 353-378, Razin et al., (Ed), Springer-Verlag, N.Y.; and Razin et al., 1991, Microbiol. Rev. 55:451-458). Recent data suggest that DNA methylation can repress gene expression directly, by inhibiting binding of transcription factors to regulatory sequences or indirectly, by signaling the binding of methylated-DNA binding factors that direct repression of gene activity (Razin et al., 1991, Microbiol. Rev. 55:451-458). It is well established that regulated changes in the pattern of DNA methylation occur during development and cellular differentiation (Razin et al., 1991, Microbiol. Rev. 55:451-458; and Brandeis et al., 1993, Bioessays 13:709-713). Importantly, the critical role of DNA methylation in differentiation has recently been demonstrated (Li et al., 1992, Cell 69:915-926; and Szyf et al., 1992, J. Biol. Chem. 267:12831-12836). The pattern of methylation is maintained by the DNA MeTase at the time of replication and the level of DNA MeTase activity and gene expression is regulated with the growth state of different primary (Szyf et al., 1985, J. Biol. Chem. 260:8653-8656) and immortal cell lines (Szyf et al., 1991, J. Bol. Chem. 266:10027-10030). This regulated expression of DNA MeTase has been suggested to be critical for preserving the pattern of methylation.
Many lines of evidence have demonstrated aberrations in the pattern of methylation in transformed cells. For example, the 5' region of the retinoblastoma (Rb) and Wilms Tumor (WT) genes is methylated in a subset of tumors, and it has been suggested that inactivation of these genes in the respective tumors resulted from methylation rather than a mutation. In addition, the short arm of chromosome 11 in certain neoplastic cells is regionally hypermethylated. Several tumor suppressor genes are thought to be clustered in that area. If the level of DNA MeTase activity is critical for maintaining the pattern of methylation as has been suggested before (Szyf, 1991, Biochem. Cell Biol. 64:764-769), one possible explanation for this observed hypermethylation is the fact that DNA MeTase is dramatically induced in many tumor cells well beyond the change in the rate of DNA synthesis. The fact that the DNA MeTase promoter is activated by the Ras-AP-l signalling pathway is consistent with the hypothesis that elevation of DNA MeTase activity and resulting hypermethylation in cancer is an effect of activation of the Ras-Jun signalling pathway.
It is clear that the pattern of methylation is established during development by sequential de novo methylation and demethylation events (Razin et al., 1991, Microbiol. Rev. 55:451-458; and Brandeis et al., 1993, Bioessays 13:709-713), the pattern being maintained in somatic cells. It is still unclear however, how methylation patterns are formed and maintained In vivo. Although a simple model has been proposed to explain the clonal inheritance of methylation patterns (Razin et al., 1980, Science 210:604-610), it does not explain how specific sites are de novo methylated or demethylated during the processes of differentiation and cellular transformation. Several lines of evidence suggest that factors, other than the state of methylation of the parental strand, are involved in targeting specific sites for methylation.
A similar mystery is how specific sites are demethylated during development and cellular transformation. One possible mechanism could be a passive loss of methylation, although an alternative hypothesis is that demethylation is accomplished by an independent enzymatic machinery.
Site specific loss of methylation is a well documented facet of vertebrate differentiation (Yisraeli and Szyf, 1985, In DNA methylation: Biochemistry and Biological significance, pp. 353-378, Razin et al., (Ed), Springer-Verlag, N.Y.; Razin et al., 1991, Microbiol. Rev. 55:451-458; and Brandeis et al., 1993, Bioessays 13:709-713). Whereas a loss of methylation could be accomplished by a passive process as described above, a series of observations have demonstrated that an active process of demethylation occurs in mammalian cells (see for example Yisraeli et al., 1986, Cell 46:409-416). Similar to de novo methylation, demethylation is directed by specific signals in the DNA sequence (Yisraeli et al., 1986, Cell 46:409-416; and see FIG. 1 herein for a model) and the probability of a site being methylated or demethylated is determined by the affinity of that site to either one of the DNA MeTase or demethylase. The affinity of each site to either enzyme is determined by the chromatin structure around the site (Szyf, 1991, Biochem. Cell Biol. 64:764-769).
In normal cells, the DNA methyltransferase is regulated and repressed, possibly by one of the tumor suppressors. An equilibrium between DNA methyl-transferase and demethylase activities maintains the methylation pattern. Methylated sites are indicated by (M) in FIG. 1. Inhibition of the repressor results in over-expression of the DNA MeTase (as indicated by the solid arrow) the genome becomes hypermethylated and tumorigenesis is initiated (tumor a). Another mechanism for up regulating the DNA methyltransferase is the activation of the Ras oncogenic pathway resulting in activation of Jun and over-expression of the DNA MeTase. However, it appears that the Ras pathway can activate the demethylase as well. The final pattern of methylation observed in this class of tumors will reflect both activities: hypermethylation (M) of sites that express low or medium affinity to the demethylase (sites 3,4,5) and hypomethylation of sites that are of high affinity but were methylated in the original cell (site number 6).
The lines of evidence that link cancer and hypermethylation are however still circumstantial. The critical question that remains to be answered is whether these changes in DNA methylation play a causal role in carcinogenesis.
The demonstration that hypermethylation correlates with carcinogenesis would be immensely useful since it could lead to methods of assessing the carcinogenic potential of cells as well as to therapeutic treatments of cancer patients. Of note, the fact that the level of DNA MeTase is limiting in mammalian cells is supported by the observation that a small elevation of cellular DNA MeTase levels by forced expression of an exogenously introduced DNA methyltransferase into NIH 3T3 cells results in a significant change in the methylation pattern (Wu et al., 1994, Proc. Natl. Acad Sci. USA 90:8891-8895).
In addition, if DNA methylation provides an important control over the state of differentiation of mammalian cells, then DNA methylation modifiers could serve as important therapeutic agents to alter the genetic program in a predictable manner and/or to restore an authentic program when it is disrupted by deregulation of DNA methylation.
Furthermore, the identification of the molecule responsible for the demethylase activity would be extremely useful for the same reasons as mentioned above, since the control of gene expression, of differentiation and cellular homeostasis appears dependent on the balance between the level of DNA MeTase and demethylase activities.